CRNN-Phoneme-Recognizer

PyTorch implementation of frame-level CRNN phoneme recognizer trained with TIMIT.

Installation

Install Python version higher than 3.7. and PyTorch version higher than 1.10.1.

If you use poetry, you can install other packages easily through the following command.

$ poetry install

If you don't use poetry, please just install packages in pyproject.toml manually.

Extract phoneme posteriorgram (PPG) with pretrained model.

You can download pretrained model from Releases, and you can save PPG through the following command.

$ python infer.py [MODEL_PATH] [AUDIO_PATH] [SAVE_PATH]

Train your own model

If you want to train your own model, you should prepare TIMIT dataset, and fix the configs/paths.yaml to your TIMIT dataset path.

Then, you can train the model through the following command.

$ python train.py [LOGGER]

LOGGER can be wandb, tensorboard, none (do not log anything for debug). If you want to fix the hyperparameter, you can fix configs/models/CRNN.yaml.